Maryland is aiming to establish itself at the forefront of regulating artificial intelligence in the healthcare sector. However, it raises an essential question: does the state fully grasp the complex landscape it is trying to oversee?
In the past year, Maryland implemented House Bill 820, a legislative effort aimed at constraining how insurers, pharmacy benefit managers, and utilization review entities apply AI in healthcare decisions. Advocates of the bill position it as a vital consumer protection initiative that seeks to prevent automated care denials and maintain the role of human clinicians.
Yet, this legislative optimism masks a more intricate reality. Modern healthcare AI functions within a vast, interconnected digital ecosystem. AI systems derive their efficacy from the data they process, which, increasingly, is not exclusively generated by healthcare professionals.
The landscape of medicine now relies on a digital framework that includes electronic health records, AI-powered scribing tools, automated coding systems, predictive analytics, and comprehensive language models integrated into clinical operations. A significant number of doctors are now utilizing AI for documentation tasks. Patient-physician interactions are recorded and summarized in real-time, with these interactions immediately fed into medical records. Insurance companies are effectively monitoring these conversations through their AI scribes — a fact that poses significant regulatory challenges.
The traditional medical chart is evolving into a collaborative product produced by both healthcare providers and machines. This shift is crucial because the AI systems employed by insurers to review prior authorizations or assess medical necessity rely heavily on the quality of the inputs they receive.
If the documentation entering the system is flawed, the outcomes will also be flawed. However, Maryland's legislation primarily addresses the use of AI on the insurer side, neglecting the overall infrastructure that generates the data these entities depend on. When an AI-based utilization management system denies care due to incomplete or incorrect documentation, determining accountability becomes increasingly complex, given that responsibility is spread across various technologies, institutions, and processes within the same healthcare workflow.
Another pressing concern is whether Maryland has the technical know-how and resources needed to properly evaluate these systems. The law calls for oversight, auditing, transparency, and human reviews, but the nuances of modern AI and algorithmic adaptation raise questions about how meaningful these audits will be.
Can state regulators adequately assess proprietary models used by insurers? Are they equipped to evaluate the training datasets, potential biases, or the behavior of algorithms? Can they discern whether negative outcomes stem from inadequate documentation by physicians, errors in AI summarization, faulty data extraction, coding issues, or algorithmic misjudgments? Alternatively, are lawmakers merely creating an illusion of regulatory oversight while being outpaced by the technology they aim to govern?
Simultaneously, as Maryland intensifies its regulation of insurer AI, healthcare institutions are fiercely deploying AI solutions to combat physician burnout, staffing shortages, and operational inefficiencies. Hospitals and doctors are already integrating AI for triaging patients, prioritizing radiology cases, predicting patient outcomes, and improving workflow.
Ironically, last year, Maryland's lawmakers also evaluated proposals that would heavily restrict physicians’ use of AI in decision-making. This approach reflects a fundamental misunderstanding of the evolution of modern healthcare.
No responsible healthcare provider is suggesting that AI should supplant human medical judgment; rather, AI serves as a tool for enhancement. For instance, radiologists using AI to pinpoint subtle imaging anomalies are not relinquishing their expertise, nor are ophthalmologists who employ AI-assisted retinal analysis. Physicians utilizing AI to process extensive patient data or identify hazardous medication interactions are practicing responsibly.
In fact, as AI technologies progress, failing to leverage these tools where beneficial may become increasingly indefensible. This raises an important issue: the discourse should never center on "AI versus physicians" but on whether AI is being applied in a responsible, transparent, and accountable manner. It also questions the regulators' ability to comprehend and effectively govern these systems.
This presents a substantial challenge for Maryland. The state cannot assert itself as a leader in technology while grappling with issues like interoperability failures, fragmented health systems, inefficient Medicaid processes, staffing shortages, and outdated IT frameworks.
Before embarking on comprehensive regulation of healthcare AI, Maryland must first confront a more unsettling question: how is AI currently functioning within its own healthcare departments? Are predictive algorithms already affecting Medicaid management, audits, reimbursement reviews, or resource distribution?
The uncomfortable reality is that the rising interest in AI regulation has drawn attention precisely because many, including policymakers, lack a full understanding of the underlying technology. This disconnect creates a risk where symbolic legislative action substitutes for effective governance.
Maryland deserves acknowledgment for recognizing the transformative potential of AI in healthcare and for addressing critical issues such as bias, transparency, and patient safety. However, to establish itself as a national leader in healthcare AI governance, the state must grapple with a larger strategic challenge: does it have the necessary expertise, infrastructure, and workforce to lead effectively?
As a significant player in healthcare and biotechnology, Maryland boasts one of the highest concentrations of physicians, research institutions, and federal health agencies in the nation. With ties to major entities such as the NIH and Johns Hopkins, Maryland is ideally situated to engage in the national discourse on healthcare AI.
However, the conversation becomes more complex. Maryland’s healthcare system, like many across the U.S., faces ongoing challenges such as shortages of healthcare professionals, fragmented IT systems, administrative delays, and increasing financial pressures on healthcare networks. The state itself also confronts fiscal challenges, leading to a critical question that seldom gets publicly examined: does Maryland truly have the financial means and technical capability to oversee intricate AI systems in healthcare as legislators envision?
Effective regulation of AI goes beyond creating task forces or issuing guidelines; it necessitates a skilled workforce knowledgeable about machine learning, algorithm auditing, health informatics, software engineering, and data governance specifically tailored for complex healthcare environments.
The adoption of AI in healthcare is accelerating rapidly, a trend that is likely to persist due to pressing economic and clinical pressures. Crafting meaningful regulatory frameworks that reflect technological realities will require collaboration among physicians, data scientists, ethicists, insurers, and healthcare regulators.
If not addressed adequately, Maryland risks only scratching the surface of healthcare AI regulation while overlooking the much deeper, consequential infrastructures at play. The key issue lies in ensuring that policymakers who shape AI regulations have a comprehensive understanding of the technology’s practical implications in clinical settings, rather than merely theoretical concepts.
